Pose Regularization Based Automatic Multi-View Face Recognition Method

Abstract

Face recognition techniques encounters difficulty in handling arbitrary poses variations. Various approaches have been worked upon for face recognition across pose variations, wherein many methods either require manual landmark annotations or assume the face poses to be known. These constraints prevent many face recognition systems from working automatically. The proposed work presents a new fully automatic multi-view face recognition method via 3D model based pose regularization, and extends existing face recognition systems into multi-view scenarios. Unlike previous pose normalization approaches, where non-frontal face images were transformed into frontal images, the proposed 3D model based pose regularization method generates synthetic target images to resemble the pose variations in query images. We should point out that generating non-frontal views from frontal face images is much easier and more accurate than recovering frontal views from non-frontal face images. This is because it is difficult to automatically detect accurate landmarks under large pose variations which are required to build a 3D face model. Additionally, since many areas of a face are significantly occluded under large pose variations, it is problematic to recover the frontal view for the occluded facial regions.The proposed work follows the view rendering based on 3D GEM but uses a simplified 3D Morphable Model [6]. Additionally, instead of aligning the synthetic target images and testing face images based on eye positions, we perform face alignment using Procrustes analysis under large pose variations. Moreover, our face matching method with blocked MLBP features provides better robustness against face illumination and expression variations. Finally, we show the expansibility of the proposed approach by replacing our MLBP based face matcher with two state- of-the-art face matching systems.